The current implementation separated logic for overfilled vs. unprocessed queues using a `move_all` flag, resulting in a somewhat convoluted loop structure.
Using a single `remaining_to_move` in the inner loop instead.

Additionally, since we are iterating to-queues from index zero, the to-index should never exceed `_num_queues`. Therefore, the modulus operation in `to_idx = (to_idx + 1) % _num_queues;` is redundant.